Nova Speak

Convert text to speech using Amazon Nova 2 Sonic via Bedrock and play through speakers.

Arguments

• $ARGUMENTS contains the text to speak, optionally with flags:
  • --voice <name> — choose a voice (default: tiffany). Options: tiffany, matthew, amy, ambre, florian, beatrice, lorenzo, greta, lennart, lupe, carlos
  • --play — auto-play through speakers (default: true)
  • --save <path> — save the WAV file to a path

Process

1. Parse arguments: Extract text and flags from $ARGUMENTS.

   • Default voice: tiffany
   • Default output: /tmp/nova-speak-output.wav

2. Generate speech: Run via Bash. Replace TEXT, VOICE, and OUTPUT:

cd ~/.claude/plugins/computer && node -e "
import { createNovaSonicBridge } from './server/services/nova-sonic.js';
import { writeFileSync } from 'fs';

const text = process.argv[1];
const voice = process.argv[2] || 'tiffany';
const outPath = process.argv[3] || '/tmp/nova-speak-output.wav';

const bridge = createNovaSonicBridge({ voice, tools: [] });
const chunks = [];
let transcript = '';

bridge.onAudio(pcm => chunks.push(Buffer.from(pcm)));
bridge.onText(t => { transcript += t; });

await bridge.connect();
bridge.sendText(text);

await new Promise(resolve => {
  let sil = 0, last = 0;
  const check = setInterval(() => {
    if (chunks.length === last) { sil++; if ((chunks.length > 0 && sil > 6) || sil > 30) { clearInterval(check); resolve(); } }
    else { sil = 0; last = chunks.length; }
  }, 500);
});

const combined = Buffer.concat(chunks);
const hdr = Buffer.alloc(44);
hdr.write('RIFF',0); hdr.writeUInt32LE(36+combined.length,4);
hdr.write('WAVE',8); hdr.write('fmt ',12); hdr.writeUInt32LE(16,16);
hdr.writeUInt16LE(1,20); hdr.writeUInt16LE(1,22);
hdr.writeUInt32LE(24000,24); hdr.writeUInt32LE(48000,28);
hdr.writeUInt16LE(2,32); hdr.writeUInt16LE(16,34);
hdr.write('data',36); hdr.writeUInt32LE(combined.length,40);
writeFileSync(outPath, Buffer.concat([hdr, combined]));

console.log(JSON.stringify({
  voice, duration: (combined.length/48000).toFixed(1),
  chunks: chunks.length, transcript, file: outPath
}));
bridge.close();
process.exit(0);
" -- "TEXT" "VOICE" "OUTPUT"

3. Play audio (unless --save specified):

afplay OUTPUT_PATH

4. Report results: Show transcript, duration, voice, and file path.

Voice Reference

Example Usage

• /computer:nova-speak Hello Captain, all systems nominal
• /computer:nova-speak --voice matthew Warp core online, shields at maximum
• /computer:nova-speak --save ~/Desktop/alert.wav Red alert, battle stations